Development of fast and accurate face recognition algorithms

Abstract

In computer vision applications face recognition has become one of the famous biometric techniques having a number of real world applications like human/computer interaction, surveillance, authentication, computer user interfaces and so on. The sensitivity of available classifiers to factors like illumination, pose variation, real world problems, uncontrolled environment and lack of robust features for classification are amongst the main problems that the researchers are facing so far. The objective of this study is to formulate a set of new approaches of face recognition, that are robust against various conditions of human face recognition including face recognition under controlled/ ideal conditions, varying lighting conditions, facial expression and pose conditions, which give higher accuracy as compared to some recent existing approaches. In general, the recognition of face images is based on their description by a set of measurable quantities called invariants that are insensitive to particular deformations and provide enough discrimination power to distinguish among various persons. Thereafter, only these invariant features are used in the recognition process and as such these extracted invariant features should be able to describe the appearance and structure of faces with only a few parameters and should efficiently represent the variations likely to occur in the face images of same or different persons. Moment based invariant features provide compact representation of the images and the same are invariant to the image rotation and the noise. Due to this quality, these are the widely used image descriptors in several applications like character recognition, image analysis and image retrieval. In this study, the performance of radial moment based approaches namely Zernike moments (ZMs) and pseudo Zernike moments (PZMs) have been analyzed for the recognition of face images.